AI Security Engineer

About the role

The AI Security Engineer will design and implement security controls for artificial‑intelligence systems, large language models (LLMs) and autonomous agents. You will protect these technologies from emerging threats such as prompt injection, jailbreaks, tool abuse and data exfiltration while collaborating closely with AI Governance and Platform Engineering teams.

Key responsibilities

• Design AI security architecture and conduct AI‑specific threat modeling.
• Implement defenses against prompt injection, jailbreaks and agent tool abuse.
• Define identity, authentication, authorization and access‑control policies for AI workloads.
• Perform AI security testing, adversarial testing and red‑team exercises.
• Carry out AI privacy reviews covering PII, prompts, outputs and logs.
• Partner with AI Governance and Platform Engineering to strengthen overall security posture.
• Communicate AI‑specific risks and vulnerabilities to security leadership.
• Identify, remediate and track emerging AI threats.
• Train delivery teams on AI security best practices and threat patterns.

Required profile

• 5+ years of experience in cybersecurity or application security, with at least 2 years focused on AI/LLM/agent security.
• Strong knowledge of LLM and agent security threats, including prompt injection, jailbreaks and data exfiltration.
• Solid foundation in IAM, network security, secure SDLC and application security.
• Experience with AI/LLM security testing and adversarial testing tools.
• Familiarity with Azure AI Foundry Prompt Shields, AWS Bedrock Guardrails and Google Vertex AI safety controls.
• Understanding of AI privacy, data‑protection and security risk management.
• Proven ability to conduct AI red‑team exercises and communicate findings to stakeholders.

Required skills

• Cybersecurity
• Application security (AppSec)
• AI/LLM security
• Identity and Access Management (IAM)
• Network security
• Secure Software Development Lifecycle (SDLC)
• AI security testing tools
• Azure AI Foundry Prompt Shields
• AWS Bedrock Guardrails
• Google Vertex AI safety controls
• NeMo Guardrails (open‑source)
• Rebuff (open‑source)